$ cat kubernetes/week/2026-09-28.log
the week in review · Sep 28 – Oct 4, 2026
Apiserver write throughput KEP opens as seven KEPs move
By RepoJournal · composed from the cited sources · human-reviewed weekly · methodology
A wave of enhancement proposals, from hugepages reservation to DRA capacity requests, reshaped the 1.38 roadmap this week.
KEP-6404: Introduce KEP (improve apiserver write throughput) kubernetes/enhancements
The new proposal targets apiserver write throughput, the path every kubectl apply, controller reconcile, and operator loop ends up in. It is early, but it is the one item on this list that scales with cluster size rather than with a single feature.
KEP-6252: Hugepages Reservation kubernetes/enhancements
The proposal adds hugepages to kubelet's --system-reserved and --kube-reserved flags, so hugepage memory can be subtracted from node allocatable the same way CPU and memory already are. Node operators running hugepage workloads currently have no way to reserve that memory out of the schedulable pool.
kep: address reviewers comments 09232026 kubernetes/enhancements
The follow-up revision drops the FeatureGate entirely: adding hugepages to --system-reserved or --kube-reserved is itself the enablement mechanism, the PR says, "similar to what FeatureGate does", and adds rollout cases to the KEP. That means no separate gate to flip at upgrade time.
KEP-4958: CSI Sidecars All in one kubernetes/enhancements
A single KEP pulls the CSI sidecar containers together, touching every deployment that currently ships separate provisioner, attacher, and resizer containers alongside its driver. Worth tracking for anyone templating storage add-ons.
KEP-6406: Allow Capacity Requests for DRA Extended Resource Mapping kubernetes/enhancements
The proposal extends capacity requests to DRA extended resource mapping, so drivers can advertise a resource and have it consumed under the same capacity semantics. Relevant to anyone already writing DRA ResourceSlices.
KEP-3953: In-place Node Resource Resize kubernetes/enhancements
The proposal would let a node's resources change in place instead of requiring a drain and recreate. Database, cache, and stateful workload operators care most; they currently absorb resize by restarting the pod.
feat: expose PVC condition last transition time kubernetes/kube-state-metrics
Exposes kube_persistentvolumeclaim_status_condition_last_transition_timestamp_seconds so Prometheus can compute how long a PVC has sat unused. Backed by KEP-5541, which defines the Unused condition; alerting on stale PVCs no longer needs a custom exporter.
fix(proxy): correctly detect explicit localhost NodePort address kubernetes/kubernetes
ContainsExplicitLoopback() treats any IPv4 loopback CIDR as an explicitly configured localhost address, so 127.0.0.2/32 gets the same handling as 127.0.0.1/32. The fix tightens detection of addresses users actually configured.
$ ls kubernetes/week/ # the briefings behind this review
Keep up with Kubernetes in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.
fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.
Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.